K. E. BANYARD AND M. R. HAYNS
416
orrelation and the Charge Distribution in Lithium Hydride
by K. E. Banyard* and M. R. H a p s Department of Physics, University of Leicester, Leicester, England
(Received February 19,1970)
Publication costs borne completely by Thti Journal of Physical Chemistry
The redistribution of charge density as a consequence of electron correlation has been examined for LiH. Two cowelated wave functions of differing complexity were analyzed. A natural orbital formulation allowed a first natural configuration to be considered as a noncorrelated limit. Thus, we minimized difficultiesassociated witlh different basis sets being used for correlated and noncorrelated calculations. In this way, by means of density difference maps, our examination of the changes in electron distribution concerned the influence of correlation introduced within the confines of each model. Both correlation maps indicated an expansion of the molecular charge cloud. The electron density increased at each nucleus and a reduction of charge occurred in the internuclear region. In a relatively extensive region around each nucleus, the density difference contours exhibited characteristics of split-shell correlation similar to those possessed by two-electron ions.
I. Introduction Wave iurictions derived from Hartree-Fock-Roothaan calculations\' have given rise to energies which are a close approximation to the true Hartree-Fock limit for many atoms and molecules. The preeminence of such independent-particle models lies in their physical and procedural simplicity. However, because the Hartree-Fock treatment itself makes no allowance for electron correlation2 other than through the Fermi hole, the calculated energy can be in error by an amount comparable with the bond dissociation energy. I n addition, electron correlation is usually necessary for a correct theoretical description of adiabatic dissociation. VC'ith the advent of sophisticated computers, correlated wave functions are now becoming available for several small systems. Clearly, it is of considerable interest to examine the influence of electron correlation on the electronrc ,&ructure. I n this respect, although the determitbatiori of various one- and two-particle expectation values can be quite illuminating they are, of necessity, integrated properties of the wave function. GonsequeiitXy, their sensitivity to localized changes in the density ma.y well be limited. Therefore, it seems worthwhile to study the molecular charge distribution itself. Dlenrrity maps and difference maps derived from noncorrelated wave functions for molecules have already h e n analyzed by several other worker^.^-^ However, in this article, we examine the changes in the charge distribution which occur as a consequence of introducing some allowance for electron correlation at different levels of approximation. The need for such an investigation hati been stressed by various workers.6z6 As an example, we have considered the ground state of LiH at its equilibrium bond length. Two different treatments have been examined. Firstly, the wave function determined by Palke and Goddard' and, secondly, the extensive configuration-interaction (CI) calculation of Bender and Davidson* based on the use of T h e Journal of Physical Chemistry, Vol. 76,No.3,1071
natural orbital^.^ Although other wave functions have been reported,l0*l1the choice of the Bender and Davidsons calculation was particularly pertinent since, at the time, it was energetically the best treatment of LiH. The result has only recently been improved upon by Boys and Handy,12 who used a transcorrelated wave function. The Palke and Coddard wave function is of interest in its own right and also represents some intermediate step toward a CI function since it leads to significantly better energies than the Nartree-Fock method and yet retains an independent-particle interpretation. 11. Difference Maps, Wave Functions,
and Electron Densities The characteristics of a density difference map are (1) C. C. J. Roothaan, Rev. Mod. Phys., 23, 69 (1951). (2) An excellent discussion of electron correlation and correlation energy has been given by P. 0. Lowdin, Advan. Chem. Phys., 2, 207 (1959). (3) See for example, T. Berlin, J . Chem. Phys., 19,208 (1951); R. F. W. Bader, J. A m e r . Chem. SOC.,86,5070 (1964); A. C. Wahl, Science, 151, 961 (1966); P. Politeer and R. E. Brown, J. Chem. Phys., 45, 451 (1966); R. F. W. Bader, W. H. Henneker, and P. E. Cade, ibid., 46, 3341 (1967); B. J. R a n d and J. Sinai, ibid., 46, 4050 (1967) ; P. E. Cade, R. F.W. Bader, W. H. Henneker, and I. Keaveny, ibid., 50, 6313 (1969); I. Cohn and K. D. Carlson, J . Phys. Chem., 73, 1356 (1969); D. B. Boyd, J. Chem. Phys , 52, 4846 (1970). (4) R. F. W. Bader, I. Keaveney, and P. E. Gade, ibid., 47, 3381 (1967). (5) C. W. Kern and M. Karplus, ibid., 40, 1374 (1964). (6) P. E. Cade and W. M. Huo, ibid., 47, 614 (1967). (7) W. E. Palke and W. A. Goddard, 111, ;bidd.,50, 4524 (1969). (8) C. F. Bender and E. R. Davidson, J. Phys. Chem., 70, 2675 (1966). (9) P. 0. Lowdin, Phys. Rev., 97, 1474 (1955); 97, 1490 (1955) ; 97, 1509 (1955); see also, P. 0. Lowdin and H. Shull, Phys. Rev., 101, 1730 (1956). (10) For example, the separated electron-pair study of LiH by D. D. Ebbing and R. C. Henderson, J. Chem. Phya., 42,2225 (1965). See also R . C. Sahni, B. C. Sawhney, and M. J. Hanley, ibid., 51, 539 (1969). (11) C. F. Bender and E. R. Davidson, ibid., 49,4222 (1968). (12) 8. F. Boys and N. C. Handy, Proc. Roy. Soc,, Ser. A , 311, 309 (1969).
ELECTRON CORRELATION AND THE CHARGE DISTRIBUTION IN LiH obviously very dependent on the nature of the reference density from which tQe differences are measured. Since the amount of correlatjon energy contained within a calculation is normally measured as the improvement in energy over the Nartree-Fock result, the true HartreeFock density would be an ideal reference function for any study of electron correlation. Unfortunately, such densities e x k t for a very few systems. Although the Har tree- Fock-Roothaan (HFR) procedure can give rise to reliable energies, Kern and Karplus5 have shown that eriergetr cally comparable results based on different basis functions can possess variations in electron density equal Ln magnitude to those which arise from properties of chemical interest. To avoid this difficulty, we reeail that the reformulation of a correlated wave furletion Q in terms of a natural spin orbital analysjsghelps to minimize the influence of the composition of the original basis set. Further, when the normalized 92 is expressed as a sum of configurations built up from the n&m-al spin orbitals, this natural expansion of the wave function is distinguished as the superposii o ~ rapid ~ ~ convergence toward tion of c o n ~ ~ u ~ aoft most the energy value E given by (!P\H/e),where H is the amiktonian of bbe system. The first natural conguration has also been foundI3 to bear a striking resemblance to the artree-Fock result in terms of energy and total ovedap. ‘The relationship between natural orbitals and Hartree-Fock orbitals has been discussed by several workers l4,I5 for example, Davidson and JonesI4 showed for HZthat the difference between such orbitals is almost equal to the f function introduced by Sinanoglu16in Lhe expansion of an N-particle wave function. The f functions represent corrections to the Nartree-Fock orbit& as a consequence of correlation and, in ganerali, th& contribution to the energy is very smalLX5 Thus, for each correlated treatment of LiH, we used the first natural configuration to determine a corresponding noncorrelated reference density. By means of a density difference function, derived by subtracting this reference density from the associated correlated density, we can examine charge redistributions which arise frcrrn infiuences of correlation effects17 contained within each wave function. We now discuss t h e wave functions considered in this analysis. For an 2V-particle sy&em, Goddard18 expressed the total w,?;ve function as
G:@(l,
2, 3, .
., N)x(l, 2, 3, . . ., N )
where CP is a spnce function and x is a product of oneelectron spin functions. G,” is a projection operator’9 determined by the properties of a permutation group SNsuch thak @(I,2 , 3, . ~,N) is antisymmetric and also a spin eigenfunction. For LiH, Palke and Goddard7 used the 61 method and wrote 2, 3, 4
=
4l(1>&%#J3(3>44(4)
417
where each 4j was a two-center molecular orbital (MO) given by
4,
=
C Cjrn@m m
The coefficients C,, were determined by the variation method. The basis set {&) was composed of Slatertype orbitals (STO’s) Is, 1s‘: 29, 3s, 2pa, 2p’a, and 3da located on the Li nucleus and Is, 28, and 2pa centered on H. The total wave function involved a linear combination of 24 possible products of four basis MO’s and the coefficients in the linear combination were determined by the operator Q”. Such a calculation corresponds to a variational valence bond approach in which electron correlation has been introduced, essentially, by means of a diff erent-orbitals-fordiff erent-spins (DODS) scheme. The CI calculation of Bender and Davidson, which involved the direct use of natura: orbitals (NO’S), accounted for 89.1% of the correlation energy for LiH. Briefly, Bender and Davidson describe their CI-NO technique as follows: an estimation of the
Table I: Total Energies and Correlation Energies for LiH, Li+, and H-Energy,
-Eoorr,
70corr
System
Method
&U
RU
LiH
CI-NOQ Glb
8.0606 8.0173 7.9873 8.0696 7.2792 7.2364 7.2799 0.5275 0.4880 0,5278
0 0733 0.0300
89.1 36.5
0.0823 0.0428
loo* 98.4
0.0435 0 0395
looe
0. 0398
100“
HFc Lif f
Exptd CX
HF
H-
f
Expt CI
HF Expt
I
OS Oe
~
99.2 0“
a Reference 8. Reference 7. e The HFR value from ref 6. Corrected for relativistic effects, see ref 6. By definition, see ref 2 and 21. Results quoted from ref 21.
d
(13) G. P. Barnett, J. Linderbrg, and H. Shull, J. Chem. Rhus., 43, 80 (1965); B. G. Anex and H. Shull, “hdolecular Orbitals in Chemistry, Physics, and Biology,” P. 0. Lowdin and B. Pullman, Ed., Academic Press, New York, N. Y., 1964,p 227; 6. V. Naearoff and J. 0 . Hirschfelder, J. Chem. Phys., 39, 715 (1963). (14) E.R. Davidson and L. L. Jones, ibid., 37, 2966 (1962). (15) 0.SinanoElu, Res. Mod. Phys., 35, 517 (1963). See also 0. Sinanozlu and D. F. Tuan, J. Chem. Phys., 38, 1740 (1963),and 0.Sinanoglu, Advan. Chem. Phys., 6 , 315 (1964). (16) 0. Sinanoglu, J. Chem. Rhus., 36, 706 (1962). (17) Within the present scheme of analysis, pair correlation terms wdl constitute the first, and dominant, correction for electron correlation interaction; see ref 15 and 16. (18) W. A. Goddard 111, Phys. Rev,, 157, 81 (1967). (19) GI’ is such that p is the representation of SN required by the spin state of the system and i indicates which of Lhej’-fold degenerate spin functions in p is to be considered. For LiH, Palke and Goddard chose i = 1 and, hence, G1 is used as a label for both the method
and the wave function. The Journal of Ph&cal Chemistry, Vol. 76,No- 9, 1971
E . . , N ) which represents an N effects contained within the framework of each treatelectron systxn can be defined as ment of LiH are given by
=p(h)
p(P)
-- N
9*(1, 2,3,
. ., N )
GP(%
= P(f):>ao
-
P(f9an
-
P(f)B,
Bd
9(1?2,3, . . . I N)d7.2d7.3 . . d7.N where, in addition, integration is carried out over all spin coordinxtes. To obtain a noncorrelated reference density, the G 1 wave function of Palke and Goddard was analyzed in terms of natural orbitals xk: the trans] the occupation numbers n k formation matrix ] A j kand are presented in Table 11. As mentioned above, the
Table 11: Transformation Matrix [ Aj k ] and Occupation Numbers nk for the N O Analysis of the GI Wave Function +i/xr .$la
+lb
cp,, +2b
nk
€01.LIW" Xl
xz
0.50632 0.51038 0.00739 0.00694 0.99969
-0.07993 -0.08054 0.54304 0.53560 0.98104
X8
-0.11185 -0,03191 1.44508 -1.42928 0.01802
x4
-2.68115 2.73686 -0.39132 0.18407 0.00137
See ref 7 for details of the molecular orbitals +i. . g _ -
CI-NO wave function of Bender and Davidson is already in R suitable form. Thus, the first natural configurationg2l3arising from each calculation for LiH was used to obtain a corresponding noncorrelated density. Such quantities, indicated by a subscript "n," are denoted by ~ ( 8 for) the ~ ~Palke and Goddard (6) treatment and ~ ( 7 when ) ~ ~ evaluated fpom the first natural configuration for the Bender and Davidson (13) wave function. For each correlated treatment of LiH, the electron (density associated with the complete wave functien7j8is similarly designated p ( ~oc) or ~ ( f ) The Journal of Phg8ical Chemistry, Vol. 76,No. 9, 1071
and GP(T)B = P ( 3 B C
for the GI and CI-NO wave functions, respectively.22
111. Discussion The electron density for LiH shown in Figure 1, i.e., P(P)~,, has the same general form as that obtained from an HFR wave f ~ n c t i o n . ~The contour map for p ( Q B c reveals the existence of two extensive regions of density of almost spherical symmetry, one associated with each nucleus. Thus, as before,* this suggests an Li+H- ionic interpretation of the density. This point is emphasized further by the steep gradient of the charge distribution behind the Li nucleus, a feature characteristic of an Li+(ls)2core. A similar behavior was found for ~ ( f ) ~ , . From the standpoint of our analysis, allowance for correlation effects contained within the wave functions of Palke and Goddard' and Bender and Davidson* ~ & P ( ? ) ~shown, respeccause density changes 1 3 p ( f ) and tively, in Figures 2a and 2b. We see that the influence of electron correlation on the molecular charge distribution reveals several features of interest. Both difference maps indicate a reduction of charge density in the mid-bond region and an increase of charge at, and immediately around, each nucleus. Further, an increase in density also occurs in a toroidal outer (20) Unless stated otherwise, all quantities are expressed in terms of Hartree atomic units. (21) K. E. Banyard and C. C. Baker, J . Chem. Phys., 51, 2680 (1969);K. E. Banyard, {bid., 48, 2121 (1968). (22) ?. locates a network of grid points (in a plane containing the ~ bond ~ . axis) relative to the Li nucleus as an arbitrary origin.
ELECTRON CORRELATION AND THE CHARGE DISTRIBUTION IN LiH _
_
p
~
_
.
_
l
(
_
.
419
-
Table 111: Electron Densities p i ? ) Evaluated Along the LiH Molecular Axis (the Internuclear Separation R is 3.015 au and I’ositive z Is Measured from the Midpoint in the Direction of the H Nucleus)
-2,2Li
Densit,y
x
P(P)Bca
0.237553 0.237632 0. ‘237550 0.338361 0.238261
P(l‘)Bn5 P(i.)ffef’
P()‘k??Lb p(~)arac
E:>
Density
x
P ( h n
0.044527 0,045056 0.042218 0.048919 0,0113751
p(P)Bne
P(%eb P(f)Fnil p(?)~p~o a
=
0.00
- 1.75
-1.50 (NLi nucleus)
-1.25
-1.00
-0.75
-0.50
-0.25
0,869109 0.870326 0.868993 0.871189 0.871422
3.38835 3.39194 3.39548 3.39215 3.39076
13.15237 13.13822 13.28063 13.19209 13.18966
3.04854 3.05118 3.05272 3.04920 3.04932
0.757670 0.758401 0.759684 0.759786 0,759721
0.205183 0.205132 0.205695 0.205576 0.205656
0.070139 0.070145 0.069519 0.070069 0.070005
0.042297 0.042565 0.040850 0.042070 0.041926
+0.25
$0.50
$0.76
$1.00
4- 1 .25
$1.50 (-H nuoleus)
4- 1.7’6
$2.00
0,057826 0.058529 0.054789 0.056488 0.056431
0.079247 0,079928 0,075996 0.077145 0.077089
0.111735 0.112051 0.109036 0.108685 0.108597
0.162695 0.162142 0.160938 0.158129 0.157863
0.246134 0.244089 0.243554 0.237925 0.237432
0.388315 0.384136 0.376792 0.367991 0.369988
0.231830 0.230011 0.230428 0.224947 0.223505
0.137023 0.136442 0.138459 0.135411 0.132801
-2.0
Derived from Bender and Davidson, ref 8.
’
Derived from Palke and Goddard, ref 7.
Derived from Cade m d Wuo, ref 6.
region around the bond axis at the position of the Li functions. The dependence of the present ~ u a n t ~ t a t i v e results on the nature of the basis set cannot, of course, nucleus. Results of a similar nature have been observed from a corresponding analysisz3for HeH+. The be assessed with absolute accuracy. Certainly, this dependence has been minimized through our use of NSO effect of electron correlation also gives rise to closed negative contours behind the Li nucleus. This implies analysis by the fact that the same basis set has been involved in determining the correlated and noncorrethat, in this region of space, a slight preferential lated estimate for the electron densities within the inerewe of chairge has occurred during molecular formabounds of each calculation examined here. tion. Such a feature is in general accord with the For each treatment of LiH, the correlated charge description by Bader. et aZ.,4 of the charge movements cloud is more diffuse throughout space than its nonassociated with ionic bonding. counterpart. A measure of this effect was Some comparison (of ~ ( f )~ ~( f~) p,~( f ) ~a o , and ~ ( 7 ) correlated ~ ~ obtained from the shift of the 0.003 contouP for p ( ~ ) from inspection of Table 111 where at positions both. along the LiH axis and perpendicular sities are given for various positions to it at the midpoint between the nuclei. For convealong t,he LiH molecular axis: x = 0 locates the midnience, each result was expressed as a percentage of the point of the internuclear separation R = 3.015 au. “size” of the molecule given by the n ~ n c o r r posie~~~~~ As a reference, vaiues for the HFR density are also tion of the contour. The Render and Davidson wave included in Table TIL We see that, in the immediate function gave relative outward shifts of the 0.803 convicinity of the nuclei, the correlated wave functions tour of 1.3% along the Li-I3 axis and 1.6% in the peryield a charge density which is larger than the appropendicular direction: the GI method gave an increase, priate noncorrelated value: also, this trend is carried due to correlation, of 1.6% and 1.8%, respectively. ~ s o n the G1 correlated results over to a c o m ~ ~ , ~between only :t rough guide. These results are, of course(?;, and the HFR .values. However, in contrast, the CIIdeally, it would be most instructive -lo consider how NO density ~ ( 1 at4 ~ the ) ~ Li~ nucleus is lower than the the spatial volume, associated with a fixed large fraction corresponding IWR result. From the comparisons of the charge, changes as a consequence of electron coravailable in Table IIII we also note that P(P)~%, rather relation. than P ( F ) ~ has ~ ~ a better overall agreement with A detailed comparison of the correlation maps can ~ ( 7 ) ~At ~this ~ .point we recall that, although the be obtained from the 6 p ( ? ) profiles given in Figure 3. P-XO calculations each use different The amount of charge redistribution (ap(?), is seen to basis sets, the JIFR and GI results are derived from be greater and spatially more extensive than 6 ~ ( 7 ) ~ . STO’s, whereas the Bender and Davidson calculation This behavior would seem to parallel the atomic situawas based on d.iptioal functions. Browne and Mattion, where the introduction of correlation by a DODS senz4comment hhat STO’s can give a better representaLion of the essentially spherical charge distributions near molecular nuclei than elliptical functions, but are (23) K. E. Banyard and C . C . Baker, Int. J . &an. Chem., 4, 431 (1970). for % description of the valence elec(24) J. C. Browne and F. A. Matsen, Phys. Rev., 135, 1227 (1964). trons. These Idmawations highlight difficulties asso(25) This contour generally contains well over 90% of the total ciated with a1197 direct interpretive comparisons beelectronic charge and hence provides some assessment of molecular tween c a ~ ~ u ~involving ~ t ~ o ~the~ use s of different basis size. The J O U Tof ~Physaoal Chem&g,
VoL 76,No. 5,1971
K. E. BANYARD AND M. R. HAYNS
420
00 0 8 0 006 0004
0 1
t -
L
2
A
3a.u.
LI nudeus ti rwckur
ST.
0002 0 000
I'
-0.002 -0.004
i
f
-0.001
-0.002 -0 003
-30
2 0 L------4----.-I
3a.u.
A
Li nucleus H nucleus
ib) Figure 2. Correlation density difference maps for LiII. (a) 6p(?)a arise? from correlation effects contained in the 61 wave function of Palke and Goddard. (b) 6 p ( ? ) ~ i s due to correlrition effects contained in the CI-NO wave function o f Bender and Davidson.
scheme overemphasizes the density change by comparison with hhe re,sults obtained from the analysis of a C I treatment.21 Figures 3b and 3c indicate that, around the LLi nucleus, each calculation produces a correlated "split-shell" effect. A similar behavior, but more diffuse, oIccurs around the H nucleus. These features are also tghown in Figure 3a; however, the decrease in charge density due to correlation effects in the internuclear region makes such an interpretation less obvious. An interesting extension of the present analysis would be ,to examine the behavior of 6 p ( ? ) for the ground state, and perhaps excited states, of LiH as we move from the separated atoms to the unitedatom limit.26 Such an examination, based on a difThe Journal of Phyakal Chemiatry, Vol. 76,No, 3,1971
I
I
I
-20
-I0
0'0
'
IO
L 20
L 3O.a"
Figure 3. Profiles of 6p(?)a and Q P ( ? ) B . (a) Along the bond axis: solid line represents Qp(&, and the broken line is I ~ ( ? ) B . (b) Values of Sp(?)a in a direction perpendicular to the bond axis a t the Li nucleus (dashed line) and the H nucleus (solid line), and ( e ) shows corresponding results for 6 p ( ? ) ~ .
ferent model, has already been reportedz3for the ground state of HeH +. Finally, a brief comment on the correlation energies2 is appropriate. For LiH, E,,, is estimated to be 0.0823 au20 where, in the absence of a Hartree-Fock energy, we have used the HFR value as our reference. The G1 method recovered only 36.5% of E,,, compared with 89.1% for the CI-NO treatment. The difference arises, essentially, from the representation of instantaneous correlation contained in the CI-NO scheme. The limitation of the G1 method was reflected, as seen, as an overdiffuseness in the correlation map. In conclusion, we note that the value of 0.0823 au for E,,, (26) Energies for LiH as a function of R have been reported recently by Bender and Davidson (ref 11) and Sahni, Sawhney, and Hanley (see ref 10). See also the separated pair calculations for LiH by E. L. Mehler, K. Ruedenberg, and D. iM.Silver, J . Chem. Phvs., 52, 1181 (1970). However, at R = 3.015 au, none of these calculations is energetically as accurate as the (31-NO wave function analyzed here.
421
8YMMETltICALLY fhJ13STITWTED h € A L O B I P H E N n S
(LiH) is quite dose to the total E,,, of 0.0833 au associated.with the Li+ and H- ions. The influence on the electron density of correlation effects inherent within each of two correlated wave functions has been investigated for LiH. A natural orbital analysis provided a first natural configuration which was used as a noncorrelated limit. Charge movements were illustrated by means of density difference maps and profile diagrams. The wave functions examined re those of Pake and Goddard and of Bender and vidson which recovered, respectively, about 36% and 89Pj, of tho?correlation energy for LiH.
Briefly, although the introduction of correlation effects caused the charge cloud to expand in both calculations, the Palke and Goddard calculation overemphasized the effect by comparison with the C I treatment. This conclusion parallels B similar situation found for atoms. Within the framework of our analysis, electron correlation increased the density a t each nucleus and also reduced the charge in the internuclear region, Further, in the vicinity of the nuclei, the charge redistributions exhibited characteristics of split-shell correlation similar to those associated with two-electron ions. Although not part of this investigation, the study suggests some general support for an Li+Hinterpretation of the bonding in LiH.
graetic Resonance Spectra and Substituent Effects for Symmetrical y Substituted Dihalobiphenyls A. €&. l'arpley, Jr.,l and J. H. Goldstein* Depwtment of Chemistry, Emory University, Atlanta, G m g k 30332 (Received August 10,1970)
I'ublk~atbn.coats assisted by Emory Unirereity
Nmr spectra have been analyzed for the twelve symmetrically substituted dihalobiphenyls. Additivity of substituent effects has been observed for coupling constants and chemical shifts. The effects of substituents an the coupling parameters have been shown to correlate quite well with substituent electronegativity, in agreernent with previous work on disubstituted benzenes. Substituent effects on the chemical shifts have been discus,sd in terms of ring current modification in the second ring and in terms of other well-known mechanisms. Downfield shifts at certain positions have been attributed to steric interactions. An inter-ring seven-bond 13-3' coupling has been observed in the case of 4,4'-difluorobiphenyl but no such inter-ring coupling was found for 3,3'- or 2,2'-difluorobiphenyl.
Introduction Considerable interest has been directed toward obtaining and interpreting the nmr parameters of substituted b e n ~ e u e s . " ~ ~Many of these studies have been concerned with additivity of substituent effects on coupling constants and chemical shifts, demonstrating the great utility of additivity values in the analysis and assignments of aromatic nmr spectra. Improvements in speotrometcr performance and the availability of high-speed, iterative computer programs for nmr spectral analysis have grleatly facilitated the study of these spectra and the resulting very precise values from these analyses have made the study of substituent effects much more reliable. Following the example of previous work,2 2 23,26--36 statistical correlations between substituent electronegativity and nmr coupling paramp
eters were established in monohalo- and dihalobenzenesZ8 where the changes in coupling values with sub(1) NDEA Fellow, 1967-1970; Tennessee Eastman Fellow, 1970-
1971. (2) (a) P.L.Corio and B. P. Dafiey, J . Amer. Chem. Soc., 78, 3048 (1956); (b) J. B. Leane and R. E. Richards, Trans. Faraday Soc., 55, 707 (1959). (3) I. Yamaguchi and N. Hayakawa, Bull. Chem. Soc. ,Pap., 33, 1128 (1960). (4) P. Diehl, Helv. C h h . Ada, 44, 829 (1961). (5) H. Spiesecke and W. G. Sohneider, J. Chem. Phya., 35, 731 (1961). (6) J. C.Schug and J. C. Deck, $bid.,37,2618 (1862). (7) J. Martin and B. P. Dailey, ibid., 37, 2594 (1962). (8) J. S. Martin and B. P. Dailey, dbid., 39, 1722 (1968). (9) 8. Castellano and C. Sun, J . ,4mer. Chem. BOG.,85, 380 (1963). (10) T.K.Wu and B. P. Dailey, J. Chern. Phya., 41, 2796 (1964). (11) 5. Castellano and J. Lorenc, J. Phya. Chern., 69, 3652 (1965). The Journal of Phy&.Chernietry,
Vol. 76,No. 8, 1971